JRRD Volume 50, Number 5, 2013

Pages 599–618

Mechanical design and performance specifications of anthropomorphic

prosthetic hands: A review

Joseph T. Belter, MS, BS;1* Jacob L. Segil;2 Aaron M. Dollar, PhD, SM, BS;1 Richard F. Weir, PhD3
1Department of Mechanical Engineering and Materials Science, Yale University, New Haven, CT; 2Department of
Mechanical Engineering, University of Colorado at Boulder, Boulder, CO; 3Biomechatronics Development Labora-
tory, Department of Veterans Affairs (VA) Eastern Colorado Healthcare System, Denver VA Medical Center, Denver,
CO; and Department of Bioengineering, College of Engineering and Applied Science, University of Colorado Denver,
Denver, CO

Abstract—In this article, we set forth a detailed analysis of the
mechanical characteristics of anthropomorphic prosthetic
hands. We report on an empirical study concerning the perfor-
mance of several commercially available myoelectric pros-
thetic hands, including the Vincent, iLimb, iLimb Pulse,
Bebionic, Bebionic v2, and Michelangelo hands. We investi-
gated the finger design and kinematics, mechanical joint cou-
pling, and actuation methods of these commercial prosthetic
hands. The empirical findings are supplemented with a compi-
lation of published data on both commercial and prototype
research prosthetic hands. We discuss numerous mechanical
design parameters by referencing examples in the literature.
Crucial design trade-offs are highlighted, including number of
actuators and hand complexity, hand weight, and grasp force.
Finally, we offer a set of rules of thumb regarding the mechani-
cal design of anthropomorphic prosthetic hands.
combination of high functionality, durability, adequate
cosmetic appearance, and affordability. We believe that,
in order to close the gap, a better understanding of the
current performance capabilities and performance needs
of anthropomorphic prostheses must be achieved and
commonly accepted measures and evaluation protocols
must be established.
Previous review articles on prosthetic hands have
been published [1–4]. Weir provides a thorough discus-
sion of prosthesis design, particularly as it relates to chal-
lenges facing people with amputation and their needs
*A portion of this article was published as Belter JT, Dollar AM. Per-
formance characteristics of anthropomorphic prosthetic hands. Pro-
ceedings of the IEEE International Conference on Rehabilitation
Robotics; 2011 Jun 29–Jul 1; Zurich, Switzerland. p. 921–27.

Key words: amputee, grasping, grippers, hands, iLimb Hand,

manipulation, Michelangelo Hand, rehabilitation, robotics, ter-
minal devices.
INTRODUCTION
Over the last two decades, there have been great
strides in the development of novel prosthetic hands and
terminal devices that take advantage of the latest techno-
logical advances, moving toward more dexterous hand
devices.* However, even state-of-the-art devices lack a
Abbreviations: ADL = activity of daily living, DC = direct
current; DIP = distal interphalange; DOF = degree of freedom;
MCP = metacarpal phalange; MIMO = multiple input, multiple
output; NBDM = nonbackdriveable mechanism; PIP = proxi-
mal interphalange; VA = Department of Veterans Affairs.
*Address all correspondence to Joseph T. Belter, MS, BS;
Yale University, Department of Mechanical Engineering
and Materials Science, 10 Hillhouse Ave, New Haven, CT
06511; 248-613-6296; fax: 203-432-6775.
Email: joseph.belter@yale.edu
http://dx.doi.org/10.1682/JRRD.2011.10.0188

599
600

JRRD, Volume 50, Number 5, 2013

from a more general level, and also reviews trends in

prosthetic hand development [1]. Extensive user studies
have also been conducted, including those by Van Lun-
teren and Van Lunteren-Gerritsen [5] and Atkins et al.
[6], that capture use and task information for numerous
prostheses from myoelectric to simple cosmetic devices
with the end goal of ranking and improving design char-
acteristics for prosthetic hands. Cipriani et al. [2] and
Biagiotti et al. [3] present summaries of the features of
current hand designs but do not discuss quantitative
details, nor how those design choices relate to grasping
and manipulation. Biddiss et al. present design priorities
as a result of a survey of upper-limb prosthesis users but
do not state the actual parameters of the devices that were
evaluated [4]. Other articles have also attempted to con-
duct performance testing on commercially available
prosthetic hands but have been limited in the number of
hands that were tested [7]. We focus on a complete set of
Figure 1.
test results, design specifications, and design justification (a) Vincent hand by Vincent Systems, (b) iLimb hand by Touch
to an extent not covered before. Additionally, we attempt Bionics, (c) iLimb Pulse by Touch Bionics, (d) Bebionic hand by
to discuss the appropriateness of design choices based on RSL Steeper, (e) Bebionic hand v2 by RSL Steeper, and
other science and survey results found in literature. (f) Michelangelo hand by Otto Bock. All hands shown without
In this article, we review the performance specifica- cosmetic glove.
tions of a wide range of commercial prosthetic hands
through presentation of our own empirical testing results
and through a review of published literature. Our analysis hands have received much media attention, the Vincent
of six commercial myoelectric anthropomorphic pros- (Vincent Systems; Weingarten, Germany) and Michelan-
thetic hands studies the latest developments in the field. gelo (Otto Bock; Duderstadt, Germany) hands are just
We then present a thorough overview of published perfor- becoming available to the public. Therefore, published
mance characteristics of prototype research hands with information on the Vincent and Michelangelo hands is
intended applications toward prosthetic design. We dis- limited. Tables 1 and 2 show the properties and charac-
cuss both the physical performance specifications (when teristics for each hand as claimed by the manufacturer or
available), as well as any justification provided by the gathered from video and secondary sources. The Sen-
developers regarding the scientific basis as to why those sorHand, developed by Otto Bock, is also listed in Tables
measures are appropriate. Finally, we present a discussion 1 and 2 as a comparison of the capabilities of a single
on potential mechanical design trade-offs in the current degree of freedom (DOF) hand with today’s multifunc-
state of the art in prosthetic terminal device development. tional hand designs. Table 1 presents a general descrip-
When appropriate, we present our own opinions on the tion of the mechanical design, while Table 2 presents the
rules of thumb for each design category discussed. grip forces, finger kinematics, and achievable grasps for
each hand. The information in Tables 1 and 2 is pre-
sented in order to provide a comparison with the empiri-
METHODS cal data collected during this study (summarized in
Tables 3–4).
Published Specifications of Commercial Hands
The six hands shown in Figure 1 represent the latest Empirical Testing of Commercial Hands
developments in commercial myoelectric hands. While Since the data provided in Tables 1 and 2 were com-
the iLimb (Touch Bionics; Livingston, United Kingdom) piled from numerous sources, we felt the need to test
and Bebionic (RSL Steeper; Leeds, United Kingdom) each hand with a uniform testing procedure to better
 601

BELTER et al. Mechanical specifications of prosthetic hands

Table 1.
Published general characteristics of commercial prosthetic hands.
Number Degrees Number Joint
Weight Overall Actuation Adaptive
Hand Developer of of of Coupling
(g) Size Method Grip
Joints Freedom Actuators Method
SensorHand Otto Bock 350–500 Glove sizes 2 1 1 DC Motor Fixed pinch No
(2011) [8–9] 7–8 1/4*
Vincent Hand Vincent — — 11 6 6 DC Motor- Linkage spanning Yes†
(2010) [10] Systems Worm Gear MCP to PIP
iLimb Touch 450–615 180–182 mm long, 11 6 5 DC Motor- Tendon linking Yes†
(2009) [11] Bionics 80–75 mm wide, Worm Gear MCP to PIP
35–41 mm thick
iLimb Pulse Touch 460–465 180–182 mm long, 11 6 5 DC Motor- Tendon linking Yes†
(2010) [11] Bionics 80–75 mm wide, Worm Gear MCP to PIP
35–45 mm thick
Bebionic RSL 495–539 198 mm long, 11 6 5 DC Motor- Linkage spanning Yes†
(2011) [12] Steeper 90 mm wide, Lead Screw MCP to PIP
50 mm thick
Bebionic v2 RSL 495–539 190–200 mm long, 11 6 5 DC Motor- Linkage spanning Yes†
(2011) [12] Steeper 84–92 mm wide, Lead Screw MCP to PIP
50 mm thick
Michelangelo Otto Bock ~420 — 6 2 2 — Cam design with No
(2012) [13] links to all
fingers
*Otto Bock glove sizes measured in inches from base of palm to tip of middle finger.
†Adaptive grip accomplished through electronic torque control, others from adaptive mechanical coupling.
DC = direct current, MCP = metacarpal phalange, PIP = proximal interphalange.

Table 2.
Published grip and kinematic characteristics of commercial prosthetic hands.
Grip Force Range of Motion Grasp Type
Precision Power Lateral MCP PIP DIP Thumb Thumb Thumb
Hand Finger/Grasp Achievable
Grasp Grasp Pinch Joints Joints Joints Flexion Circumduction Circumduction
Speed Grasps
(N) (N) (N) (°) (°) (°) (°) (°) Axis
SensorHand NA 100 NA 0–70* NA NA 0–70* NA None Up to Power
(2011) [8–9] 300 mm/s
at tip
Vincent Hand — — — 0–90* 0–100* NA — — Parallel with — Power, precision,
(2010) [10] wrist axis lateral, hook,
finger-point
iLimb 10.8 — 17–19.6 0–90* 0–90* ~20 0–60* 0–95* Parallel with 200 mm/s Power, precision,
(2009) [11] wrist axis lateral, hook,
finger-point
iLimb Pulse — 136 — 0–90* 0–90* ~20 0–60* 0–95* Parallel with 1.2 s (power grasp) Power, precision,
(2010) [11] wrist axis lateral, hook,
finger-point
Bebionic 34 75 15 0–90 10–90 ~20 — 0–68 Parallel with 1.9 s (power grasp), Power, precision,
(2011) [12] (tripod) wrist axis 0.8 s (tripod grasp), lateral, hook,
1.5–1.7 s (key grasp) finger-point
Bebionic v2 34 75 15 0–90* 0–90* ~20 — 0–68 Parallel with 0.9 s (power grasp), Power, precision,
(2011) [12] (tripod) wrist axis 0.4 s (tripod grasp), lateral, hook,
0.9 s (key grasp) finger-point
Michelangelo 70 NA 60 0–35* NA NA — — Compound — Opposition,
(2012) [13] axis lateral, and
neutral mode
*Estimated based on images and videos.
DIP = distal interphalange, MCP = metacarpal phalange, NA = not applicable, PIP = proximal interphalange.
602

JRRD, Volume 50, Number 5, 2013

Table 3. RESULTS
Measured commercial entire hand system weight (g).
Small Finger Large Finger Entire System
Hand Weight
Weight Weight Weight
Vincent 29–31 35–37 — The commercial hand weights are presented based on
iLimb 48 52 — the weight of the entire system required to be carried by
iLimb Pulse — — 539* the user. For the iLimb Pulse and Bebionic v2 hands, this
Bebionic v2 — — 527 includes the battery, controller, two force sensing resistors
Michelangelo — — 746* (used to simulate electromyography electrodes), and the
*Includes protective sleeve. distal side of the Otto Bock Electronic quick-disconnect
wrist unit. The Michelangelo hand entire system weight
includes the hand with protective sleeve (498 g), a much
Table 4.
Motor specifications for commercial hands. larger battery (143 g), controller (14 g), and an Axon
Gear Ratio, Rotation wrist adapter (91 g). The Vincent fingers have
Motor three different-sized distal segment attachments that allow
Hand Motor to
Type
MCP Joint the same base to be used for the three large fingers of the
Vincent Maxon 1017 — hand. The distal segment is illustrated in Figure 2(a).
iLimb Maxon RE 10 4.5 V 1.5 1600:1 Each of the end segments weighs 2 to 4 g.
W Part # 118394
iLimb Pulse Maxon RE 10 4.5 V 1.5 1600:1 Actuation Method
W Part # 118394
Bebionic Custom Linear Drive —
from Reliance Preci- Finger Kinematics
sion Mechatronics Unlike human hands, five of the six commercial
Bebionic v2 Custom Linear Drive — hands tested feature a proximal joint, similar to the
from Reliance Preci- human metacarpal phalange (MCP), and a single distal
sion Mechatronics joint that takes the form of both the human proximal
Michelangelo Custom Modified — interphalange (PIP), and distal interphalange (DIP). An
Maxon EC45 additional feature on the distal finger segment gives the
MCP = metacarpal phalange.
look of the DIP joint in the iLimb and Bebionic fingers.
The Michelangelo fingers consist of a single finger seg-
compare and discuss details regarding the hand designs. ment actuated only at a single point like that of the
Our analysis of each of these hands allows us to discuss human MCP joint and seen in Figure 2(d).
the hands side by side in a more consistent manner. Addi- Instead of actuating each joint of the fingers indepen-
tionally, experimental analysis allowed us to make obser- dently, the fingers of the iLimb, Vincent, Bebionic, and
vations regarding the kinematics of each hand that would Bebionic v2 fingers have a fixed movement relative to
have been unobtainable otherwise. each other. Figure 2 illustrates the mechanism used to
define the fixed relationship between the joint motions.
Tested Commercial Hand Systems Although these hands use a form of a four-bar linkage,
Elements of the six commercial prosthetic hands each has a distinct method of coupling the motion of the
shown in Figure 1 were acquired and tested to measure PIP to the motion of the MCP joint. The Vincent finger
their performance characteristics. The iLimb Pulse, Bebi- (Figure 2(a)) uses two externally located wire links
onic, Bebionic v2, and Michelangelo hands were tested in mounted between the finger base and the distal link. This
a fully assembled hand configuration. The iLimb Prodig- four-bar linkage mechanism, as illustrated in Figure 2(a)
its (same fingers and control system as standard iLimb) (bottom), is common among fully actuated robotic finger
and Vincent Hand finger performance characteristics were designs. The iLimb and iLimb Pulse hands use a tendon
determined through testing of a set of four fingers con- system in which a loop of fibrous cable is wrapped around
nected to a nonstandard palm mount using the same con- a bearing surface mounted to the finger base. The distal
troller and battery as the original entire hand system. end of the tendon loop is attached to the distal link and

Figure 2.
Commercial finger images (top) and kinematic models of finger
joint coupling mechanism (bottom). (a) Vincent (Vincent Sys-
tems), (b) iLimb and iLimb Pulse (Touch Bionics), (c) Bebionic
v2 and Bebionic (RSL Steeper), and (d) Michelangelo (Otto
Bock). θ1 = angle of metacarpal phalange joint, θ2 = angle of
proximal interphalange joint.
guided up the finger by two small rollers, as seen in Figure
2(b) (bottom). The rollers help to control the moment arm
created by the tendon across the PIP joint. The Bebionic
and Bebionic v2 use a similar four-bar linkage system to
the Vincent finger but use a single plastic connecting rod
between the base and the distal link that runs directly down
the center of the proximal finger segment.
The PIP to DIP joint coupling ratio was obtained using
video analysis of the finger motion during a single finger
flexion/extension motion. The joint angles were obtained
using a MATLAB (MathWorks; Nattick, Massachusetts)
script with the zero angle positions recorded as illustrated
in Figure 2(a–c). Figure 3 shows the results, including a
linear fit plotted for the entire data set for each finger. The
Vincent finger had a linear coupling ratio of PIP angle
change to MCP angle change of 1.27. The plateau in the
603
BELTER et al. Mechanical specifications of prosthetic hands
Figure 3.
Vincent (Vincent Systems), iLimb (Touch Bionics), and Bebionic
v2 (RSL Steeper) hands feature linear relationship between
metacarpal phalange (MCP) and proximal interphalange (PIP)
joints during flexion/extension motion.
Vincent finger plot from 125° to 130° of PIP motion corre-
sponded to the hard limits of travel for the distal link while
the proximal link continued in a flexion motion and was
therefore not included in the linear fit. The iLimb and
Bebionic v2 hands had similar PIP angle change to MCP
angle change ratios of 1.09 and 1.14, respectively. The PIP
to MCP ratio controls how the fingers wrap around objects
of different size. In human hand motion, the MCP to PIP
motion ratio is different during grasp acquisition motions
for objects of different size [14].
Motor Type and Packaging
Because of the extreme packaging constraints
imposed by the hand size, small motors that incorporate
high gear reductions are placed in either the proximal
phalanx (as in the iLimb, iLimb Pulse, and Vincent hands
shown in Figure 4(b–c)) or, if available, in the palm (as
in the Bebionic, Bebionic v2, and Michelangelo hands
shown in Figure 4(a)). Table 4 lists the motors and gear-
heads used for each commercial prosthetic hand. The
Vincent, iLimb, and iLimb Pulse hands all use Maxon
DC series 10 motors (Maxon Motor; Sachseln, Switzer-
land) [15]. The iLimb and iLimb Pulse use a Maxon GP
10A with metal 64:1 three-stage planetary gear reduction
before entering into a 1:1 set of bevel gears and finally a
25:1 custom worm drive located at the base of the
604

JRRD, Volume 50, Number 5, 2013

Figure 4.
(a) Drive mechanism of Michelangelo hand (Otto Bock). Center drive element controls flexion of all four fingers and thumb. Second
motor (which actuates against bronze worm gear) independently controls abduction/adduction of thumb. (b) Vincent finger motor (Vin-
cent Systems) is housed in proximal phalange and rotates worm against fixed worm gear to flex finger. (c) iLimb finger (Touch Bionics)
is actuated in same manner as Vincent finger but uses set of bevel gears between motor and worm drive. MCP = metacarpal phalange.

fingers. Based on the motor data sheets published by the Table 5.

motor manufacturer, the maximum torque that can be Finger flexion/extension speed.
Average Speed Number of Standard
generated about the MCP joint for the iLimb is 0.98 Nm Finger
(°/s) Trials Deviation
(assuming 30% efficiency for the worm drive, 73% effi- Vincent Large 103.3 2 3.0
ciency for the planetary transmission, and 92% efficiency (ring, middle, and index)
for the bevel gear set) [15]. The Bebionic and Bebionic Vincent Small (little) 87.9 2 5.1
v2 hands use a custom linear drive developed by Reli- iLimb Large (middle) 81.8 4 3.3
ance Precision Mechatronics (Huddersfield, United iLimb Med (index/ring) 95.3 2 3.4
iLimb Small (little) 95.4 2 2.6
Kingdom). The Michelangelo hand uses one large cus- iLimb Pulse Thumb 110.6 4 4.1
tom modified brushless Maxon EC45 motor housed iLimb Pulse Large 60.5 4 1.8
directly in the center of the palm to control flexion/exten- (index, middle)
sion of all five fingers and one smaller motor (type iLimb Pulse Med (ring) 74.3 4 2.8
unknown) in the proximal portion of the thumb to control iLimb Pulse Small (little) 82.2 4 4.0
Bebionic Thumb 36.6 16 7.7
thumb abduction/adduction. Figure 4(a) shows the novel
Bebionic Large 45.8 8 2.2
central drive system that actuates all five digits simulta- (ring, middle, and index)
neously through several linkage mechanisms. Bebionic Small (little) 37.8 8 5.2
Bebionic v2 Large 96.4 2 0.4
Finger Flexion Speed (ring, middle, and index)
Michelangelo (index) 86.9 4 2.8
Individual finger flexion/extension speeds were meas- med = medium.
ured about the MCP joint using an externally mounted
potentiometer. The calibrated time-based voltage data
were used to determine the average finger speed over the fingers were given a 100 percent command signal to the
entire flexion/extension motion (0°–102° for Vincent, 0°– controller for the entire duration of motion.
91° for iLimb, 0°–60° for iLimb Pulse, 0°–60.6° for Bebi-
onic, and 0°–35° for the Michelangelo). The data pre- Grip Force
sented in Table 5 show the individual finger speeds for the The individual finger forces were measured using a
six hands. The full hand finger speed data correspond with calibrated load-cell. For the individual finger measure-
the speed of the fingers when all fingers are flexed or ments, the load-cell was placed at the finger tip of each
extended simultaneously in free air. During each run, the finger with the finger in the fully extended position. The
 605

BELTER et al. Mechanical specifications of prosthetic hands

entire hand was commanded to close at full power and The grasp force was measured on the commercial
then released. Although there is a force peak at impact, the hands using pinch meters for precision grasps and a grip
constant holding force is the value presented in Tables 6 dynamometer for lateral grasp and power grasps. Each
and 7. The Vincent and iLimb Pulse hands use an addi- device was calibrated over the range of loads experienced
tional pulse mode to increase the individual finger holding
during each test. The individual finger holding force was
force. After a set period of time of motor stall, quick pulses
of power are sent to the motor. The effect is to “ratchet” the not measured for the Michelangelo hand since all digits
system to a higher capable holding force than was previ- are actuated by a central drive as opposed to a single
ously experienced. The pulse mode increased the holding drive per finger in the other commercial hands.
force of an individual finger in the Vincent hand by
an average of 69.5 percent and in the iLimb Pulse by an Compliance
average of 91.5 percent. However, the pulse mode greatly Each hand design features a mechanical element that
reduces battery life. allows for a certain level of compliance in the flexion
direction. This type of feature helps to prevent the fingers
Table 6. from breaking under any inadvertent contact, forcing the
Individual finger holding force at tip. fingers to close. The Vincent finger features a unique
Force Number of Standard bend in the links connecting the base and distal segment.
Finger
(N) Trials Deviation
Vincent Large 4.82 or 8.44* 14 or 8* 0.8 or 1.3 * The bend allows it to act like a series of elastic elements
(ring, middle, and index) and enables the distal link to move under excess force
Vincent Small (little) 3.00 2 0.1 with the MCP joint remaining fixed. The iLimb and
iLimb Large (middle) 7.66 2 0.2
iLimb Pulse hands use a simple spring and tendon drive
iLimb Med (index/ring) 5.39 4 0.1
iLimb Small (little) 5.17 2 0.1
that allows the distal link to flex inward independent of
iLimb Pulse Med (index) 4.15 or 6.54* 1 — the MCP joint. The Bebionic and Bebionic v2 are the
iLimb Pulse Large (middle) 3.09 or 6.24* 2 or 2* 0.7 or 0.4* only hands that allow for compliance in both the MCP
iLimb Pulse Med (ring) 6.43 or 11.18* 2 or 2* 0 or 0.3* and DIP joints. Although they are rigidly coupled to each
iLimb Pulse Small (little) 4.09 or 8.56* 2 or 2* 0.1 or 0*
Bebionic (index) 12.47 1 —
other, the actuator is connected to the proximal link
Bebionic (middle) 12.25 2 1.0 through a pinned slot. If the finger is forced in the flexion
Bebionic (ring) 12.53 2 1.1 direction, the pin simply rides up the slot, allowing the
Bebionic Small (little) 16.11 2 0.2 MCP and DIP joints to flex inwards. Figure 2(a) shows
Bebionic v2 Large 14.5 2 1.2 the curved linkages of the Vincent finger and the pinned
(ring, middle, and index)
*Holdingforce after pulse mode.
actuation slot of the Bebionic v2 finger. Figure 5 shows
med = medium. the direction of compliance and actuation linkage of the

Table 7.
Overall grasp holding force during grasp postures.
Lateral Grasp Palmer Grasp Power Grasp
Hand Total Force Number Standard Total Force Number Standard Total Force Number of Standard
(N) of Trials Deviation (N) of Trials Deviation (N) Trials Deviation
iLimb Pulse 17.04 or 3 or 3* 2.8 or 2.0* 10.82 or 2 or 2* 0.5 or 0.3* Large Grip: Large Grip: Large Grip: —
32.10 * 17.11 * 65.25 or 71.44* 1 or 2 or 4.0*
Small Grip: Small Grip: 1* Small Grip: —*
50.81*
Bebionic 17.61 1 — 29.47 1 — 77.37 1 —
Bebionic v2 16.4 4 3.2 22.53 4 1.5 62.4 6 10.3
Michelangelo 50.84 4 3.1 78.14 8 4.4 Grasp Type Grasp Type Grasp Type
Unachievable Unachievable Unachievable
*Holding force after pulse mode.
— indicates no standard deviation because only one trial performed.
606
JRRD, Volume 50, Number 5, 2013
Figure 5.
Flexion compliance in Bebionic v2 fingers (RSL Steeper) is
accomplished by slot connection between proximal phalanx
and linear actuator.
Bebionic v2 fingers. The Michelangelo hand has direct
coupling of the actuator to the finger MCP motion
through compliant linkages. The linkages for the index
and middle fingers are made of a compliant plastic link-
age and the linkages for the ring and little finger are
extension springs.
Thumb Design and Position
A variety of thumb designs and positions are used in
the hands tested. The iLimb, iLimb Pulse, Bebionic, and
Bebionic v2 thumbs have actuated distal joints (i.e., MCP
and PIP), while the circumduction joint can be positioned
in multiple states manually (the Vincent hand tested did not
include a thumb). The two positions for the Bebionic v2
circumduction joint are shown in Figure 6 (dotted lines).
The relationship between the rotation axis of the thumb
and the main axis of the wrist is a critical design parameter
since it determines the trajectory of the thumb and there-
fore the kinematics of the functional grasps. The Michelan-
gelo hand has a complex thumb joint that is prepositioned
by a small motor prior to performing a grasp. This small
motor changes the path that the thumb will take when the
main motor actuates to close the hand either in a palmer or
lateral grasp. The thumb of the Michelangelo hand also has
a natural-looking rest position.
DISCUSSION
Survey of State-of-the-Art Research Hands and
Discussion of Mechanical Design Parameters
The empirical findings described in the “Results”
section are supplemented with the following survey of
state-of-the-art research hands in this section. Here, we
discuss the key features of prosthesis design with the end
goal of collecting comments made in the literature that
would support or motivate a particular design specifica-
tion. To aid in the discussion, we also present a review of
13 prototype research hands. The selection of these 13
hands was based on a specific distinguishing feature of
the design that warrants discussion in light of improving
and determining the ideal performance characteristics of
commercial prosthetic hands. It is important to make the
distinction between prototype research hands and com-
mercial hands since many prototype hands are developed
as a means to demonstrate a particular feature and not to
prove an entire hand system and therefore cannot be
compared 1:1 with the entire system parameters of com-
mercial hands.
Physical Properties (Weight and Size)
Hand weight. The human hand has an average
weight of 400 g [16] (wrist disarticulation, and not
including the forearm extrinsic muscles) or 0.6 percent of
the total body weight for men and 0.5 percent for women
[17]. However, prosthetic terminal devices of similar
weight have been described by users as being too heavy
[18]. Since the forces from the device are borne by the
soft tissue instead of the skeleton, the perceived weight in
the terminal device is increased. Although researchers are
currently working to alleviate attachment problems
through the use of customized socket design and osseoin-
tegrated attachment mechanisms [19], the weight of the
prosthesis is a key contributor to interface discomforts
and use fatigue. A recent Internet survey of myoelectric
prosthetic users concluded that 79 percent considered
their device “too heavy” [18]. Also, in a similar survey,
Biddiss et al. found that users rated the weight of the
device as 70 on a scale of 0 (not important) to 100 (most
important) in regards to the design priorities of prosthetic
hands [4]. In addition to the overall weight of the device,
the weight distribution affects the perceived weight of the
system. For this reason, it is desired to move heavier
components including actuators and batteries as proximal
as possible within the prosthesis.

Figure 6.
Illustration of circumduction axis location for Bebionic v2 thumb (RSL Steeper) (shown from bottom view).
Tables 1 and 8 show the weight of both current pros-
thetic hands and research hands designed for use in pros-
theses. A range of 350 to 615 g is seen in current
commercial prosthetics and 350 to 2,200 g in research-
based hands. Data presented in the tables are based on
values published by the various research groups and do
not reflect a consistent comparison of weight. For some
hands, the entire actuation and control system including
batteries and wrist attachment is included in the total
weight of the hand. Others only consider the weight of
the hand itself and not the external computing or power
sources for operation.
Within the prosthetics community, no set specifica-
tion exists for the maximum weight of the prosthesis. The
only agreed upon specification is to minimize weight in
general. Ultimately, the weight will relate to the required
size and capabilities of the hand. According to Pons et
al., an adult-sized prosthetic hand should weigh less than
400 g [20]. Kay and Rakic have set a requirement that the
entire hand including cosmetic glove should remain
under 370 g [21], while other groups, including Light and
Chappell [22] and Vinet et al. [23], believe a 500 g
weight limit is appropriate.
Hand size. For an anthropomorphic prosthesis, it is
natural for the envelope of the hand to replicate the size
and shape that is natural to the user. All of the myoelectric
607
BELTER et al. Mechanical specifications of prosthetic hands
hands, shown in Table 1, are designed to be covered with
a silicone glove to enhance the cosmetic appearance of
the prosthesis. Since prosthetic hands are sized according
to human hand measurements (and commonly based on
direct measurements of the patient’s able hand), the pros-
thetic hand structure, including cosmetic covering, should
have a length between 180 and 198 mm and a width of 75
to 90 mm to match normal human hand size [11].
Actuation Properties
Finger kinematics. Anatomically correct finger kine-
matics are a goal in mechanical design of prosthetic hands.
However, there is a trade-off between anatomical correct-
ness and robustness, weight, complexity, and cost. In many
of the hands reviewed in this article, there are more joints
than number of actuators. Often, numerous joints will be
coupled to act as a single compound motion where only the
actuator position, for example, must be known to determine
the position of all joints that are coupled together. A distinct
set of movements that can be described by a single parame-
ter is considered a single DOF. The four fingers of the
MANUS-Hand (collaboration between Consejo Superior
de Investigaciones Científicas, Argana del Rey, Spain;
Ketholiek Universiteit Leuven, Belgium; Centro de Recu-
peracion de Minusvalidos Fisicos, IMSERSO, Spain; Alor-
man Advanced Medical Technologies Ltd, Israel; and
608

JRRD, Volume 50, Number 5, 2013

Table 8.
Published general characteristics of 13 research hands with applications in prosthetics.
Number Degrees Number
Weight Overall Actuation Joint Coupling Adaptive
Hand Developer of of of
(g) Size Method Method Grip
Joints Freedom Actuators
TBM Hand* University of 280 146 mm long, 15 6 1 DC Motor with Compliant Yes
(1999) [24] Toronto 65 mm wide, Linear Ball springs
25 mm thick Screw
Remedi Hand University of 400 Similar to human 14 6 6 DC Motor Coupled MCP, No
(2000) [22] Southampton hand (Maxon) DIP, PIP
RTR II ARTS/Mitech 350 — 9 9 2 DC Motors Tendon and Yes
(2002) [25] Laboratories free-spinning
(Pisa Italy) pulleys
MANUS-Hand Spain/Belgium/ 1200 — 9 3 2 Brushless DC Fixed coupling No†
(2004) [20] Israel Motors of MCP, PIP,
and DIP
DLR/HIT I DLR German Space 2,200 1.5× human hand 17 13 13 Brushless DC 1:1 coupling of No
(2004) [26] Agency, Harbin Motors with two distal flexion
Institute of Planetary Drive joints
Technology
DLR/HIT II DLR German 1,500 Human hand size 20 15 15 Brushless DC 1:1 coupling of No
(2008) [26–27] Space Agency Motors with two distal flexion
Harmonic joints
Drive
UB Hand 3 University of — Human hand size 18 15 16 HiTec Servos PIP and DIP No
(2005) [28] Bologna, Italy coupled in ring,
little, and thumb
UNB Hand University of New — Size 7.5 10 5 3† DC Motors Fixed coupling of Yes
(2010) [29–30] Brunswick (MicroMo PIP to MCP
1724)
FluidHand III Forschungszen- 400 Similar to human 8 8 1 pump, Pressurized Distributed Yes
(2009) [31] trum Karlsuhe hand 5 valves fluid pressure
GmbH (KIT)
Smarthand ARTS Laboratory, 520 12 mm longer and 16 16 4 DC Motors Tendon/spring Yes
(2009) [2,32] Pontedera Italy 8 mm thicker (Faulhaber) based
than 50% male
Keio Hand Keio University, 730 320 mm length 15 15 1 Ultrasonic Motor Single tendon Yes
(2008) [33] Yokohama, Japan (with motor), for each finger
120 mm fingers
Vanderbilt Hand Vanderbilt 580 190 mm long, 16 16 5 Brushed DC Single cable for Yes
(2009) [34] University 330 mm with Servomotors each finger
motors, 75 mm mounted in
wide Forearm
LO/SH University of — — 8 4 2 DC Motors Wiffle tree along Yes
Southampton Southampton finger
Hand (2001) [35]
*Designed for children.
†Two degrees of freedom
of thumb controlled through single motor.
DC = direct current, DIP = distal interphalange, MCP = metacarpal phalange, PIP = proximal interphalange.

Advanced Material Technologies N.V., Belgium) [20] are
considered one DOF (despite having 8 joints) since they are
directly coupled to one another. This is an example of a rig-
idly coupled hand. Another way of coupling is through
adaptive underactuation, in which a single actuator controls
a number of independent DOFs [36]. In this sense, the sin-
gle actuator parameter cannot be used to describe the posi-
tion of the joints since they are dependent on the contact
state of each finger link with the object. These mechanisms
are considered adaptive because, when they are used in a
hand, they allow multiple links of the fingers to passively
adapt to the shape and location of an object with a single
actuator [37–38]. Examples of adaptive finger designs in
prosthetics include a single tendon routed across multiple

joints, such as in the Vanderbilt hand (Center for Intelligent
Mechatronics, Vanderbilt University; Nashville, Tennessee)
[34] and RTR-II (ARTS/Mitech Labs, Scuola Superiore
Sant’Anna; Pisa, Italy, and Centro INAIL RTR; Viareggio,
Italy) [25], or the compliant spring connections used in the
TBM hand (Department of Mechanical and Industrial Engi-
neering and Institute of Biometerials and Biomedical Engi-
neering, University of Toronto; Toronto, Canada, and
Rehabilitation Engineering Department, Bloorview Mac-
Millian Center; Toronto, Canada) [24] and Smarthand
(ARTS Laboratory, Sculuola Superiore Sant’Anna; Pont-
edera, Italy) [2,32].
Tables 2 and 9 show the range of finger motion for
both commercial and research hands. For commercial
hands, the PIP and MCP joints exhibit similar ranges of
motion to the human hand. The DIP joint, however, is
usually fixed at 20°.
Thumb kinematics. The thumb design in an anthro-
pomorphic prosthetic hand is critical since the thumb
accounts for arguably 40 percent of the entire functional-
ity of the human hand [39]. In most of the prosthetic
hands described in this article, the thumb is actuated in
flexion/extension (simple closing or opening) and along
the circumduction axis. The circumduction rotation of the
thumb is the movement required to alternate between a
lateral grasp and a power or precision grasp. An analysis
of human hand kinematics shows an average circumduc-
tion motion of 90.2°, which is achieved through a combi-
nation of three joints at the base of the thumb [40]. As
can be seen in Tables 2 and 9, the circumduction axis of
current hands is not always oriented parallel with the
wrist rotation axis. By angling this axis ventrally or dor-
sally, thumb flexion and circumduction rotation can be
jointly approximated in a single DOF. This can be benefi-
cial to achieve desired hand openings and a more anthro-
pomorphic motion for precision, power, and lateral grasp
patterns while keeping complexity low. The coupling can
also help the timing of the grasp if all of the fingers are
actuated simultaneously. Further discussion of the role of
the thumb circumduction axis can be found in other
reviews [1,21,23,40].
Type of actuator and drive mechanism. The most
common actuator used in prosthetics today, excluding a
body-powered harness, is a direct current (DC) motor.
These motors are small and lightweight and can be pack-
aged in the hand. Brushed DC motors are more commonly
used in prosthetic hands because of their ease of control.
Brushless DC motors provide higher torque-to-weight
609
BELTER et al. Mechanical specifications of prosthetic hands
capabilities but require more complex motor control
schemes. Brushless motors typically include sensors that
can provide additional position feedback. Moreover, as
control electronics continue to shrink in size, brushless
DC motors will likely become the dominant motor choice.
All DC motors naturally produce excessive speed and
insufficient torque for use in prosthetic devices. There-
fore, drive reductions are necessary to reduce the speed
and increase the torque provided by the actuator [1].
In order to reduce the speed and increase the limited
torque from these motors, gearing, lead screws, and even
harmonic drives may be used. The iLimb and Vincent
hands package a single motor and gear train in the proximal
phalange of each finger. The FluidHand III (Forschung-
szentrum Karlsruhe GmbH; Eggenstein-Leopoldshafen,
Germany) uses a small DC motor to drive a small hydraulic
pump housed within the palm of the hand [29]. Five inde-
pendent valves then transmit pressure to bellows located
at each joint. The advantage of using a pressure-based sys-
tem is the compliance associated with each finger joint,
which allows the system to survive sudden impacts. Many
of the hands incorporate nonbackdriveable mechanisms
(NBDMs) between the motor and the flexion of the fingers.
NBDMs allow the finger to maintain high grip forces
(assisted by compliance in the mechanism) without contin-
ued current draw from the battery. The most common
NBDMs include lead screws, worm drives, and roller
clutches. See Weir [1] and Controzzi et al. [41] for addi-
tional information regarding NBDMs.
Grip force. Most activities of daily living (ADLs)
require fast speed and low grip force (e.g., typing, gestur-
ing). However, tasks that require low speeds and high
grip force occur often enough that a prosthetic hand must
enable the user to perform such tasks (e.g., opening door
with handle, unscrewing jar lid).
The grip force able to be exerted by a hand on an
object is largely a function of the hand posture, object
geometry, and transmission method. In particular, pros-
thetic hands like the Hosmer Hook (Hosmer; Campbell,
California), SensorHand [8–9], and TBM Hand [24] will
exhibit different grasp forces depending on the size of the
object. The necessary grasp force to maintain an object
within a particular grasp is also difficult to predict
because it is largely dependent on the friction between
the fingers of the hand and the object, the number of con-
tact points, the relative locations of contact, and the
object geometry and mass properties. In a precision
grasp, the human hand can exert an average of 95.6 N of
force [1]. In power grasps, the forces can reach up to
610

JRRD, Volume 50, Number 5, 2013

Table 9.
Published grip and kinematic characteristics of 13 research hands.
Grip Force Range of Motion Grasp Type

Hand Precision Power MCP PIP DIP Thumb Thumb

Thumb Finger/Grasp
Grasp Grasp Joints Joints Joints Circumduction Circumduction Achievable Grasps
Flexion Speed
(N) (N) (°) (°) (°) (°) Axis
TBM Hand 14.0 — 0–90 10–50 10–50 45 to +70 Parallel with — 90° in 4–5 s Power, precision,
(1999) [24] (from perpendicular wrist axis lateral, hook,
to palm plane) tripod
Remedi Hand 9.2 — 0–81 — — — 10° toward thumb — Full thumb Power, precision,
(2000) [22] from wrist axis* motion in lateral, hook,
2.5 s tripod, finger-
point, counting
RTR II — — — — — 0 to 90* 45° toward little — — Power, precision,
(2002) [25] finger from wrist lateral
axis*
MANUS-Hand 60.0 — 0–45* 0–55* 0–70* 10 to 85* 45° toward thumb — Full grasp in Power, precision,
(2004) [20] from wrist axis* 1.2 s lateral, hook
DLR/HIT I 7.0 — — — — 0 to 90* Parallel with wrist — 180°/s Power, precision,
(2004) [26] axis lateral, hook,
tripod, finger-
point, counting
DLR/HIT II 10.0 — 0–90 0–90 0–90 20 to 20† None Same as — Power, precision,
(2008) [26–27] fingers lateral, hook,
tripod, finger-
point, counting
UB Hand 3 6.8 — 0–90 0–90 0–90 — Fixed rotation but Same as Full closure Power, precision,
(2005) [28] finger adduction/ fingers in 0.36 s lateral, hook,
abduction tripod, finger-
point, counting
UNB Hand — — 0–90 0–90 — 0 to 120 Parallel with wrist PIP joint Power, precision,
(2010) [29–30] axis only lateral, hook,
tripod, finger-
point
FluidHand III 45.0 — 0–90* 0–80* ~35 0 to 90* 10° toward little — 1 s closing Power, precision,
(2009) [31] finger from wrist time lateral, hook,
axis* finger-point
Smarthand — — 0–90 — — 0 to 120 40° toward little — 1.4 s for full Power, precision,
(2009) [2,32] finger from wrist open or close, lateral, hook, tri-
axis* thumb flexion pod, finger-point,
in 0.67 s counting‡
Keio Hand — 37 — — — 90 None — Full closure Power, precision
(2008) [33] in 0.8 s

Vanderbilt Hand 20.0 80 0–90 0–90 0–90 10 to 80 15° toward little — 225°/s, 0.4 s Power, precision,
(2009) [34] finger from wrist to close lateral, hook,
axis* finger-point
LO/SH 45.0 — — — — — — PIP joint Full close Precision/tripod
Southampton Hand only <1.2 s
(2001) [35]
*Estimated based on images and videos.
†Abduction/adduction of thumb but not rotationabout circumduction axis.
‡No independent control of fingers 3–5.
DIP = distal interphalange, MCP = metacarpal phalange, PIP = proximal interphalange.

400 N [1]. According to Heckathorne [42], a grip force of suggest a minimum grip force of 45 N for prosthetic
only 68 N is required to carry out ADLs [42]. Vinet et al. hands for practical use [23].

Tables 2 and 9 show the published grasp force mea-
surements in three grasp configurations for common
prosthetic and research hands. The more dexterous robot
hands, such as the DLR/HIT II (Institute of Robotics and
Mechatronics German Aerospace Center DLR; Wessling,
Germany, and Robot Research Institute of Harbin Insti-
tute of Technology; Harbin, China) [26–27] and the UB
Hand 3 (University of Bologna; Bologna, Italy) [28] have
a lower grip force than the simpler SensorHand [8–9] and
MANUS-hand [20]. This trend is due to a mechanical
design trade-off between complexity and strength when
constrained by size. Figure 7 shows the relationship
between the number of actuators and the published grip
force during a palmar/precision grasp for multiple pros-
thetic hands and research hands. These results vary
widely because of differences in actuator size, transmis-
sion ratio, and mechanism friction.
Grasp speed. According to Pylatiuk et al., 100 per-
cent of females, 76 percent of males, and 50 percent of
children surveyed (4, 26, and 7 subjects, respectively)
would describe the speed of their myoelectric prosthesis
to be “too slow” [18]. Although the human hand can
exhibit finger flexion speeds of 2,290 °/s, the typical
speeds for everyday pick and place tasks is 172 to 200 °/s
[1,42]. Tables 2 and 9 show the published grasp speeds of
numerous prosthetic hands. Since the data compiled in
Figure 7.
Comparison between number of actuators and palmer/preci-
sion grip force in prosthetic and research hands.
611
BELTER et al. Mechanical specifications of prosthetic hands
these tables are based on published information, there are
numerous ways the speeds have been described. What is
of most concern to the end user, however, is the amount of
time it takes to acquire an object in different possible
grasp configurations. Some groups, therefore, present
grasp speed as a measure of time to open or close the
hand. Presenting hand speed data in terms of total time to
acquire an object is problematic since the metric is depen-
dent on the size and shape of the object. The finger flexion
speeds for the hands surveyed in this article ranged from
20 °/s (TBM hand, 4–5 s to close grasp) to 225 °/s
(Vanderbilt hand, 0.4 s to close). Tözeren suggests that a
0.8 s closing time is sufficient for prosthetic hands [17].
Dechev et al. states a slower 1.0 to 1.5 s closing time is
adequate for conducting ADLs [24]. In fact, closing
speeds that are too fast can be a substantial negative
because many myoelectric control schemes rely on the
user to stop the hand at the right closing position while it’s
moving (i.e., no direct position control); excessive closing
speed makes that substantially more difficult.
Achievable grasps. The typical ADLs conducted by
an amputee can be accomplished using a finite set of pre-
defined grasps. These grasp patterns include power (used
in 35% of ADLs), precision (used in 30% of ADLs), lat-
eral (used in 20% of ADLs), hook, tripod, and finger
point [2]. Some researchers consider certain gesturing to
be important (e.g., finger counting) [2]. The full range of
distinct grasp types for the able hand is greater than 30,
and detailed descriptions of these can be found in Cut-
kosky [43]. Tables 2 and 9 show the ability of each hand
studied within this article to form these grasp patterns
without considering contact forces with the object. In
order for a hand to accomplish all seven grasping patterns
(six standard grasps plus finger counting), each individ-
ual finger flexion motion must be controlled with an
independent actuator that is not coupled to the other fin-
gers. However, removing the requirement for finger
counting can reduce this to a smaller number, particularly
if external interaction is permitted, such as a common
feature for thumb circumduction axis to be passive and
changed by the user, as is the case with the TBM, iLimb,
and Bebionic hands. A passive thumb mechanism
requires an external force to maneuver the thumb into
distinct postures and cannot be moved by the device.
Many hands, such as the TBM hand, attempt to accom-
plish as many patterns as possible with fewer than five
individual actuators. This hand uses a single actuator
612
JRRD, Volume 50, Number 5, 2013
with passive movement of the thumb circumduction axis
to accomplish five of the seven common grasp patterns.
Durability/Cycles of Use
The average myoelectric prosthetic hand user will
wear his or her device in excess of 8 h per day [18]. Any
device must therefore be robust enough to withstand pro-
longed use and comfortable enough for the user to wear
for this amount of time. The mechanical design of the
hand must consider the trade-off between durability and
robustness with weight, expense, and size. Compliant
components like conforming fingertip/palmar pads, com-
pliant actuators designs, collapsible linkage systems, etc.,
add robustness and function while not adding overdue
complexity. In general, increasing robustness typically
reduces complexity (i.e., number of DOFs, number of
components) and increases size and weight.
According to Zheng et al., between 2,500 and 3,000
grasping motions of the unaffected dominant hand may
be performed over an 8 h period during the work day
[44]. A prosthetic device will typically undergo 120
grasping motions per day [5]. The predicted grasps of the
prosthesis is lower than the able hand since a reduction in
functionality will likely result in less frequent use. A
report by Vinet et al., intended to put forth specifications
for electromechanical hands, claimed that prostheses
should withstand 300,000 grasping cycles and maintain
all of its original functionality [23]. Given the daily
expected number of cycles described previously, this
would put the lifetime of the device at about 6 yr. It is
suggested that current devices last in excess of 500,000
grasp cycles between routine servicing.
Discussion of Hand Design Trade-offs
Because of the strict limitations on size and weight
for practical prosthetic hands, the performance trade-offs
between various design options must be addressed by the
designer. The information presented in this article may be
used to benchmark performance and compare various
prosthetic and research hand designs.
Number of Actuators Versus Hand Complexity
Based on the data presented in Tables 1 and 2, a
comparison can be made between the weight of each
hand and the number of actuators used. Figure 8 shows
that although there may be an increase in weight of the
hand associated with the number of actuators, the cou-
pling of multiple joints to one or two motors can still
Figure 8.
Relationship between weight of surveyed hands and number of
actuators.
greatly increase the weight of the hand because of the
increase in mechanical complexity to implement the cou-
pling, as illustrated with the Keio (System Design and
Management and Integrated Design Engineering, Keio
University; Yokohama, Japan) [33] and MANUS-Hand
[16]. Figure 9 shows that the total number of joints in the
hand is strongly correlated with the weight of the hand,
regardless of coupling methods. Figure 10 shows the
relationship between the total number of joints and the
number of actuators for the hands presented in this arti-
cle. Hands lying on the dashed line have a single motor
for each joint of the hand with no coupling between
joints. The hands that fall to the right of the dashed line
indicate that they contain some form of coupling between
joints. A large group of research hands are contained in
the range of 15 to 20 joints, which approaches the num-
ber of joints in the human hand (~30).
The choice of the number of actuators in the hand has
traditionally been based on the type of tasks that hand is
designed to achieve. In theory, a grasping hand can be
designed with 2 DOFs (actuators) since grasping is a low
dimensional task [45] requiring a minimum of 1 to
2 DOFs to execute all functional grasps (lateral, palmar,
power). A more dexterous hand capable of a high number
of grasp configurations and individual finger motions typi-
cally requires a high number of actuators regardless of the
type of coupling used in the hand. However, clever use of
coupling strategies like whiffle tree designs or differential

Figure 9.
Distribution of weights of prosthetic and research hands plotted
against number of joints in each hand.
Figure 10.
Comparison between number of actuators and number of joints
in prosthetic and research hands.
mechanisms can enable both a wide range of grasping and
in-hand manipulation performance [28,46–47].
Hand Weight Versus Grasp Force
Figure 11 compares the hand weight and precision
grasp strength of all the hands studied in this article. The
SensorHand has the highest precision grip force to weight
ratio of all the hands studied, and the DLR/HIT II hand
613
BELTER et al. Mechanical specifications of prosthetic hands
Figure 11.
Distribution of hand weight compared with amount of grip force
hand can exert in palmer/precision grasp configuration.
has the lowest. Besides these outliers, both the research
and commercial prosthetic hands have similar precision
grip force to weight ratios. The reason for these outliers
may be the specialization of the hand. The SensorHand
has a single DOF that is controlled with a mechanism that
increases the grip force as the hand closes. The DLR/HIT
II hand is designed specifically for dexterity without con-
cern for practical tasks requiring higher grip forces.
Multiple Inputs Versus Multiple Outputs
In general, the design of a prosthetic hand must solve
a multiple input, multiple output (MIMO) problem.
Inherently, there are too few inputs (i.e., myoelectric con-
trol sites) compared with the desired number of outputs
(i.e., DOFs of the mechanical device). The mechanical
design of the devices, and in particular establishing effec-
tive schemes to couple joints/fingers to one another either
mechanically or in software/control, can help solve the
MIMO problem by reducing the number of DOFs that
need to be directly controlled by the user. The focus on
effective coupling is being widely used in modern pros-
theses and will likely continue to be refined in the future.
Advanced prostheses with more than one or two
actuators suffer from a lack of sufficient control channels
as input to the device from the user since a typical myo-
electric control system can only decipher one or two con-
trol signals. This prompts designers to prioritize specific
movements of the hand and couple motions together that
614
JRRD, Volume 50, Number 5, 2013
are used more frequently. The fingers are typically cou-
pled to a single actuator in either a direct rigid coupling
or an adaptive underactuated method.
Underactuated hands present several design advan-
tages compared with rigidly linked hands. Adaptive fin-
gers can interact with objects over more locations and
thereby distribute the grasping force over more contacts.
Also, the mechanism can take a greater range of configu-
rations for the same number of actuators. Kargov et al.
concluded that, while the contact forces are higher when
using a fixed coupling fingered prosthesis, the joint
torques of adaptive fingered hands are comparable with
the joint torques of human hands [48].
Although some commonly used prosthetic hands
allow for adaptability in grasping, a study by Bergman et
al. in 1992 claimed that a conventional nonadaptive pros-
thesis showed “significantly better results” regarding
width of grip, force of grip, and scores in a standardized
grip function test when compared with a similar adaptive
prosthesis [49].
Research Versus Commercial Prostheses
Research hands are typically “one-off” prototypes
developed as “proof of concept” devices related to a
novel design approach or as a platform to study a pros-
thetic control method. Often times, the researchers that
produce them are not focused on or not interested in
many of the details related to whether they might be com-
mercially viable. Therefore, they are often not designed
for many of the issues relevant to successful commercial
products, including robustness, weight, ease of mass pro-
duction, and cost. Along these lines, it is clear through
the successes/failures of various commercial devices and
durability history of research hands that there are gener-
ally design trade-offs between the complexity of the hand
(often including the number of DOFs) and the durability/
robustness of the hand. Accordingly, the least robust
hands are commonly the highly dexterous and compli-
cated hands, typically developed for research purposes.
Features and Specifications Versus Practical Functionality
Although numerous prosthetic hands can be com-
pared using their relevant features and specifications, the
actual goal is to create a practical device for users. Even
with the technology available today, the most frequently
used prosthetic terminal devices are still the split hook
type devices (such as the Hosmer Hook). The reasons for
this include its practicality and ease of use for accom-
plishing typical tasks, high durability, light weight, and
low cost. If we simply directly compare hand features
and specifications, the split hook would never be consid-
ered the best device, but from a practical sense, it has
proven to remain the most common choice. Therefore,
although comparison of hand features and specification
can tell us about the similarities and differences in hand
designs, it may not tell us the full story about the level of
potential benefit to end users.
Design Iterations and Continual Improvement
The design of commercial and robotic hands is a con-
tinual process that requires numerous versions and itera-
tions to perfect. The hands featured in this article
represent the versions or iterations of the design that were
available to the authors at the time of publication. Multi-
ple publications are made regarding the development and
testing of robotic hands that often contained improved
performance measurements with later versions. Please
refer to the cited documents for details regarding the
exact version of the hand specifications presented in this
article.
CONCLUSIONS
The information presented in this article serves as a
compilation of empirical and published hand characteris-
tics and performance measures. Within this article, we
focused on the mechanical characteristics of hands with-
out treatment of sensing, controls, electronics, and power
requirements and techniques.
Since a hand, like any other tool, has many uses, suffi-
cient performance for one application might not be appro-
priate for another. It is therefore difficult to establish exact
mechanical and performance requirements. Ultimately,
the selection of hand characteristics and specification is a
choice between trade-offs in complexity, dexterity (e.g.,
achievable grasps), weight, and control methods. Further-
more, each of these measures are subject to the patients’
exact needs, including the nature and level of their ampu-
tation, as well as level of activity, professional needs, and
others. The entire prosthesis must work as a system to
facilitate the accomplishment of tasks.
A set of clinical standards for performance, including
techniques for evaluating anthropomorphic hand designs,
would be beneficial and is planned for future work by the
authors. It is clear from this review that the current

performance standards used by hand designers span a
wide enough range that many would be considered unac-
ceptable in a practical device. Working toward a common
set of standards (or range of standards) would help maxi-
mize the likelihood that the extensive research efforts in
this area might be implemented in a successful commer-
cial device that will improve the lives of the population it
is meant to serve.
With that said, the authors feel confident in prescrib-
ing the following rules of thumb for the mechanical
design of anthropomorphic prosthetic hands.
• The total weight of the prosthesis (including mecha-
nism, glove, electronics, etc.) should be below 500 g.
A lighter prosthetic hand is particularly better for peo-
ple with high-level amputation because of power and
weight constraints of the entire prosthetic arm.
• Simple and robust finger kinematic designs are pre-
ferred at this time over anatomically correct finger
designs.
• Powered adduction of the thumb is highly desirable
since it allows for active posture control such as switch-
ing from lateral prehension to palmar prehension.
• The use of brushless motors instead of brushed motors
is preferred because of performance versus weight
considerations.
• A maximum pinch force at the finger tip of 65 N dur-
ing palmar prehension is recommended.
• 230 °/s should be achieved by a high-performing pros-
thesis, while 115 °/s is a minimal acceptable speed.
• Compliance in the mechanical design of a prosthetic
hand can be achieved in various ways (conforming
fingertip/palmar pads, compliant actuators designs,
collapsible linkage systems, compliant joints, etc.)
and is highly recommended by the authors.
• Highly functional grasping hands should be designed
with a low number of actuators with transmissions
that allow for all functional grasping postures.
The rules of thumb listed here focus on the mechani-
cal design criteria that the authors are confident in pre-
scribing as a universal opinion, and therefore not all
mechanical design criteria discussed earlier in this study
are addressed. However, the list provides a thorough
foundation upon which mechanical designers of pros-
thetic hands can reference.
615
BELTER et al. Mechanical specifications of prosthetic hands
ACKNOWLEDGMENTS
Author Contributions:
Background research and review: J. T. Belter.
Study concept and design: A. M. Dollar, R. F. Weir.
Acquisition of data: J. T. Belter, J. L. Segil, A. M. Dollar, R. F. Weir.
Analysis and interpretation of data: J. T. Belter, J. L. Segil.
Drafting of manuscript: J. T. Belter, J. L. Segil, R. F. Weir.
Critical revision of manuscript for important intellectual content:
A. M. Dollar, R. F. Weir.
Administrative, technical, or material support: R. F. Weir.
Financial Disclosures: The authors have declared that no competing
interests exist.
Funding/Support: This material was based on work supported by the
Gustavus and Louise Pfeiffer Research Foundation; Department of Vet-
erans Affairs (VA) Rehabilitation Research and Development Service,
administered through the VA Colorado Eastern Healthcare System; and
a U.S. Department of Defense award (grant W81XWH-10–1-0921).
REFERENCES
1. Weir RF. Design of artificial arms and hands for prosthetic
applications. In: Kutz M, editor. Standard handbook of bio-
medical engineering and design. New York (NY):
McGraw-Hill; 2003. p. 32.1–32.59.
2. Cipriani C, Controzzi M, Carrozza MC. Objectives, criteria
and methods for the design of the SmartHand transradial
prosthesis. Robotica. 2010;28(6):919–27.
http://dx.doi.org/10.1017/S0263574709990750
3. Biagiotti L, Lotti F, Melchiorri C, Vassura G. How far is the
human hand? A review on anthropomorphic robotic end-
effectors. Bologna (Italy): University of Bologna; 2008.
4. Biddiss E, Beaton D, Chau T. Consumer design priorities
for upper limb prosthetics. Disabil Rehabil Assist Technol.
2007;2(6):346–57. [PMID: 19263565]
http://dx.doi.org/10.1080/17483100701714733
5. Van Lunteren T, Van Lunteren-Gerritsen E. In search of
design specifications for arm prostheses. In: Stassen HG,
Sheridan TB, Van Lunteren T, editors. Perspectives on the
human controller: Essays in honor of Henk G. Stassen.
Boca Raton (FL): CRC Press; 1997.
6. Atkins DJ, Heard DC, Donovan WH. Epidemiologic over-
view of individuals with upper-limb loss and their reported
research priorities. J Prosthet Orthot. 1996;8(1).
7. Waryck B. Comparison of two myoelectric multi-articulating
prosthetic hands. Proceedings of the 2011 MEC Symposium;
2011 Aug 14–19; New Brunswick, Canada.
8. System electric hand size 7 [Internet]. Duderstadt (Ger-
many): Otto Bock; 2013. Available from:
http://www.ottobock.com/cps/rde/xchg/ob_us_en/hs.xsl/
6952.html
616
JRRD, Volume 50, Number 5, 2013
9. SensorHand speed [Internet]. Duderstadt (Germany): Otto
Bock; 2013. Available from:
http://www.ottobock.com/cps/rde/xchg/ob_com_en/hs.xsl/
3652.html
10. VINCENT hand [Internet]. Weingarten (Germany): Vin-
cent Systems; 2013. Available from:
http://handprothese.de/vincent-hand/
11. Touch Bionics web site [Internet]. Mansfield (MA): Touch
Bionics Inc; 2013. Available from:
http://www.touchbionics.com/
12. RSL Steeper web site [Internet]. Leeds (United Kingdom):
RSL Steeper; 2013. Available from: http://rslsteeper.com/
13. Michelangelo operation manual. Duderstadt (Germany):
Otto Bock; 2012.
14. Ingram JN, Körding KP, Howard IS, Wolpert DM. The sta-
tistics of natural hand movements. Exp Brain Res. 2008;
188(2):223–36. [PMID:18369608]
http://dx.doi.org/10.1007/s00221-008-1355-3
15. Maxon Motor. Motor data [Internet]. Sachseln (Switzer-
land): Maxon Motor; 2013. Available from:
http://shop.maxonmotor.com/ishop/app
16. Chandler RF, Clauser DE, McMconville JT, Reynolds HM,
Young JW. Investigation of inertial properties of the human
hand. Washington (DC): U.S. Department of Transporta-
tion; 1975 Mar. Report No. DOT HS-801 430.
17. Tözeren A. Human body dynamics: Classical mechanic
and human movement. New York (NY): Springer; 2000.
18. Pylatiuk C, Schulz S, Doderlein L. Results of an Internet
survey of myoelectric prosthetic hand users. Prosthet
Orthot Int. 2007;31(4):362–70. [PMID:18050007]
http://dx.doi.org/10.1080/03093640601061265
19. Jönsson S, Caine-Winterberger K, Brånemark R. Osseoin-
tegration amputation prostheses on the upper limbs: meth-
ods, prosthetics and rehabilitation. Prosthet Orthot Int.
2011;35(2):190–200. [PMID:21697201]
http://dx.doi.org/10.1177/0309364611409003
20. Pons JL, Rocon E, Ceres R, Reynaerts D, Saro B, Levin S,
Van Moorleghem W. The MANUS-HAND* Dextrous
Robotics Upper Limb Prothesis: Mechanical and manipula-
tion aspects. Auton Robots. 2004;16:143–63.
http://dx.doi.org/10.1023/B:AURO.0000016862.38337.f1
21. Kay HW, Rakic M. Specifications for electromechanical
hands. Proceedings of the 4th International Symposium on
the External Control of Human Extremities; 1972 Aug 28–
Sep 2; Belgrade, Yugoslavia. p. 137–55.
22. Light CM, Chappell PH. Development of a lightweight and
adaptable multiple-axis hand prosthesis. Med Eng Phys.
2000;22(10):679–84. [PMID:11334753]
http://dx.doi.org/10.1016/S1350-4533(01)00017-0
23. Vinet R, Lozac’h Y, Beaudry N, Drouin G. Design method-
ology for a multifunctional hand prosthesis. J Rehabil Res
Dev. 1995;32(4):316–24. [PMID:8770796]
24. Dechev N, Cleghorn WL, Naumann S. Multiple finger,
passive adaptive grasp prosthetic hand. Mechanism Mach
Theory. 2001;36(10):1157–73.
http://dx.doi.org/10.1016/S0094-114X(01)00035-0
25. Massa B, Roccella S, Carrozza MC, Dario P. Design and
development of an underactuated prosthetic hand. Proceed-
ings of the 2002 IEEE International Conference on Robot-
ics and Automation; 2002 May 11–15; Washington, DC.
p. 3374–79.
26. Liu H, Wu K, Meusel P, Seitz N, Hirzinger G, Jin MH, Liu
YW, Fan SW, Lan T, Chen ZP. Multisensory five-fingered
dexterous hand: The DLR/HIT Hand II. Proceedings of the
2008 IEEE/RSJ International Conference on Intelligent
Robots and System; 2008 Sep 22–26; Nice, France.
p. 3692–97.
27. DRL Institute of Robotics and Mechatronics. Datasheet of
DLR Hand II [Internet]. Cologne (Germany): DLR; 2011.
Available from:
http://www.dlr.de/rm/en/desktopdefault.aspx/tabid-3802/
6102_read-8922/
28. Lotti F, Tiezzi P, Vassura G, Biagiotti L, Palli G, Melchiorri
C. Development of UB Hand 3: Early results. Proceedings
of the 2005 IEEE International Conference on Robotics
and Automation; 2005 Apr; Barcelona, Spain. p. 4488–93.
29. Losier Y, Clawson A, Wilson A, Scheme E, Englehard K,
Kyberg P, Hudgins B. An overview of the UNB hand sys-
tem. Proceeding of the 2011 Myoelectric Controls/Powered
Prosthetics Symposium; 2011; New Brunswick, Canada.
30. Clawson A, Segil J, Jones B, Losier Y, Kyberg PJ, Weir R.
Mechanical design of a multifunction hand prosthesis sys-
tem—The UNB Hand. Proceedings of the 13th ISPO
World Congress; 2010 May 10–15; Leipzig, Germany.
31. Gaiser IN, Pylatiuk C, Schulz S, Kargov A, Oberle R, Wer-
ner T. The FLUIDHAND III: A multifunctional prosthetic
hand. Am Acad Orthot Prosthet. 2009;21(2):91–96.
32. Cipriani C, Controzzi M, Carrozza MC. Mechanical design
of a transradial cybernetic hand. Proceedings of the 2008
IEEE/RSJ International Conference on Intelligent Robots
and Systems; 2008 Sep 22–26; Nice, France. p. 576–81.
33. Kamikawa Y, Maeno T. Underactuated five-finger pros-
thetic hand inspired by grasping force distribution of
humans. Proceedings of the 2008 IEEE/RSJ International
Conference on Intelligent Robots and Systems; 2008 Sep
22–26; Nice, France. p. 717–22.
34. Dalley SA, Wiste TE, Withrow TJ, Goldfarb M. Design of
a multifunctional anthropomorphic prosthetic hand with
extrinsic actuation. IEEE/ASME Transactions on Mecha-
tronics. 2009;14(6):699–706.
http://dx.doi.org/10.1109/TMECH.2009.2033113
35. Kyberd PJ, Light C, Chappell PH, Nightingale JM, What-
ley D, Evans M. The design of anthropomorphic prosthetic
hands: A study of the Southampton Hand. Robotica. 2001;

19:595–600.
http://dx.doi.org/10.1017/S0263574701003538
36. Birglen L, Laliberte T, Gosselin C. Underactuated robotic
hands. Berlin (Germany): Springer; 2008.
37. Dollar AM, Howe RD. The highly adaptive SDM hand:
Design and performance evaluation. Int J Robot Res. 2010;
29(5):585–97.
http://dx.doi.org/10.1177/0278364909360852
38. Crowder RM, Dubey VN, Chappell PH, Whatley DR. A
multi-fingered end effector for unstructured environments.
Proceedings of the 1999 IEEE International Conference on
Robotics and Automation; 1999 May 10–15; Detroit,
Michigan. p. 3038–43.
39. Ouellette EA, McAuliffe JA, Caneiro R. Partial-hand
amputations: Surgical principles. In: Bowker JH, editor.
Atlas of limb prosthetics: Surgical, prosthetic, and rehabili-
tation principles. St. Louis (MO): Mosby Year Book; 1992.
40. Coert JH, van Dijke HG, Hovius SE, Snijders CJ, Meek MF.
Quantifying thumb rotation during circumduction utilizing a
video technique. J Orthop Res. 2003;21(6):1151–55.
[PMID:14554232]
http://dx.doi.org/10.1016/S0736-0266(03)00114-1
41. Controzzi M, Cipriani C, Carrozza MC. Miniaturized non-
back-drivable mechanism for robotic applications. Mech
Mach Theory. 2010;45(10):1395–1406.
http://dx.doi.org/10.1016/j.mechmachtheory.2010.05.008
42. Heckathorne CW. Upper-limb prosthetics: Components for
adult externally powered systems. In: Bowker JH, editor.
Atlas of limb prosthetics: Surgical, prosthetic, and rehabili-
tation principles. St. Louis (MO): Mosby Year Book; 1992.
43. Cutkosky MR. On grasp choice, grasp models, and the
design of hand for manufacturing tasks. IEEE Trans Robot
Autom. 1989;5(3):269–79.
http://dx.doi.org/10.1109/70.34763
44. Zheng JZ, De La Rosa S, Dollar AM. An investigation of
grasp type and frequency in daily household and machine
shop tasks. Proceedings of the 2011 IEEE International
617
BELTER et al. Mechanical specifications of prosthetic hands
Conference on Robotics and Automation; 2011 May 9–13;
Shanghai, China. p. 4169–75.
45. Santello M, Flanders M, Soechting JF. Postural hand syner-
gies for tool use. J Neurosci. 1998;18(23):10105–15.
[PMID:9822764]
46. Dollar AM, Howe RD. The SDM hand as a prosthetic ter-
minal device: A feasibility study. Proceedings of the IEEE
10th International Conference on Rehabilitation Robotics;
2007 Jun 13–15; Noordwijk, the Netherlands.
47. Odhner LU, Dollar AM. Dexterous manipulation with
underactuated elastic hands. Proceedings of the 2011 IEEE
International Conference on Robotics and Automation;
2011 May 9–13; Shanghai, China.
48. Kargov A, Pylatiuk C, Martin J, Schulz S, Döderlein L. A
comparison of the grip force distribution in natural hands and
in prosthetic hands. Disabil Rehabil. 2004;26(12):705–11.
[PMID:15204492]
http://dx.doi.org/10.1080/09638280410001704278
49. Bergman K, Ornholmer L, Zackrisson K, Thyberg M.
Functional benefit of an adaptive myoelectric prosthetic
hand compared to a conventional myoelectric hand. Pros-
thet Orthot Int. 1992;16(1):32–37. [PMID:1584641]
Submitted for publication October 7, 2011. Accepted in
revised form November 29, 2012.
This article and any supplementary material should be
cited as follows:
Belter JT, Segil JL, Dollar AM, Weir RF. Mechanical
design and performance specifications of anthropomor-
phic prosthetic hands: A review. J Rehabil Res Dev.
2013;50(5):599–618.
http://dx.doi.org/10.1682/JRRD.2011.10.0188